The Role of Attention on Viewpoint-Invariant Object Recognition
Researchers in the field of visual perception have dedicated a great deal of effort to understanding how humans recognise known objects from novel viewpoints (often referred to as shape constancy). This research has produced a variety of theories—some that emphasise the use of invariant representations, others that emphasise alignment processes used in conjunction with viewpoint-specific representations. Although researchers disagree on the specifics of the representations and processes used during human object recognition, most agree that achieving shape constancy is computationally expensive—that is, it requires work. If it is assumed that attention provides the necessary resources for these computations, these theories suggest that recognition with attention should be qualitatively different from recognition without attention. Specifically, recognition with attention should be more invariant with viewpoint than recognition without attention. We recently reported a series of experiments, in which we used a response-time priming paradigm in which attention and viewpoint were manipulated, that showed attention is necessary for generating a representation of shape that is invariant with left-right reflection. We are now reporting new experiments showing that shape representation activated without attention is not completely view-specific. These experiments demonstrate that the automatic shape representation is invariant with the size and location of an image in the visual field. The results are reported in the context of a recent model proposed by Hummel and Stankiewicz ( Attention and Performance16 in press), as well as in the context of other models of human object recognition that make explicit predictions about the role of attention in generating a viewpoint-invariant representation of object shape.